Validation of a questionnaire for detecting potential physical abuse in children

VALIDATION OF A QUESTIONNAIRE FOR DETECTING POTENTIAL PHYSICAL ABUSE IN CHILDREN. S. Mu~oz, M. Aracena, L. Bustos, P. Lorca, F. Roman, D. Streiner. CEU, Universidad de la Frontera, Temuco, Chile. Objective: This research studied the validity of a questionnaire for detecting potential physical abuse in children from a low socio-economic status in Temuco, Chile, compared to expert group assessment. Design: Validity study using cross-sectional data. Set t ing: Low socio-economic families with children under 15 years of age living in the city of Temuco, Chile. Participants: A non probabilistic sample of 522 parents, having at least one child 14 years of age or younger, was used to assess the validity of the instrument. The sample was restricted to those parents whose children had received attention in either a health care, an educational or justice public services of the city. The sample was stratified in three groups: physical abuse (128), at risk of maltreatment (184), and no maltreatment (210). M a i n Outcome Measure(s): Physical abuse assessed by a group of experts (physicians, psychologists, social workers and statisticians) was used as a gold standard. A questionnaire concerning social support, child's expectations, aduh emotional stability, and interview-derived past history data was used to construct the instrumel-a to be validated. Stat is t ical M e t h o d s : Items and domain selection was done through the determination of the internal consistency of items in each domain, and test-retest reliability on domains. Construct validity was evaluated by using confirmatory factor analysis. Stepwise backward logistic regression was used fi*r extreme group validation. Results: Cronbach's alphas ranged from 0.78 to 0.95 indicating that selected items in each domain have high internal consistency. Test-retest reliability of the final domains had a minimum of 0.74. Sensitivity and specificity of the instrument are 62.5% and 95.65% respectively; the final logistic regression model correctly classifies 90.0% of the subjects. Conclusions: To predict child physical abuse in low socio-economic status families is a very complex and muhidimensional problem. The high specificity and the moderate sensitivity of the instrument (which is expected when the prevalence is low) allow detection of children at low risk of being physically abused. The moderate sensitivity is due partly because it Was not possible to interview parents known to be abusers. The main strength of the instrument is the fact that it considers both domains internationally recognized as important as well as idiosyncratic characteristics of the Chilean population. CHILDREN'S HEALTH